Implemented and trained 6 NFL drive outcome prediction Neural Net models on NFL
data to create coefficients for sports betting companies. Each model is responsible for a specific type of drives in NFL

Given raspberry cameras set in a shoe store the video pipeline takes care of processing videos and
detecting all the interactions of customers with shoes. The interactions are supposed to then be used
for statistical analysis of shoes’ popularity. Processing includes detection and classification of shoes,
persons, person faces and emotions and save the interactions for further analysis and statistics.
Furthermore, each shelf (2 shelves per store) has 2 raspberry cameras set for it. The first camera
looking at the shelf and the second looking from the shelf and used to detect mainly but not
exclusively customer face and emotions. 

Scraped 10+ million company front pages from LinkedIn (Java, Ubuntu, Selenium, DOM)

Created a data pipeline to quickly digest and synchronize each NBA game’s tracking and play-by-play data into structured units - “Possessions”. 

This framework deals with big and raw data and is easily maintained. It also allows the developer to collect specific data(such as shot probability data, post play and ball_screen datasets) for further analysis.